Multilayered Extended Semantic Networks (MultiNet) have been developed along the general line of semantic networks (SN) for the semantic representation of large stocks of natural language information. They allow for a very differentiated meaning representation of natural language expressions and an adequate modelling of cognitive structures. MultiNet is permanently used for a fully formal semantic characterization of lexemes in a large computational lexicon and as a semantic interlingua in natural language interfaces and question-answering systems. Apart from the structural information defined by relations and functions over the nodes of the SN (which is a feature common to all SN), MultiNet is characterized by embedding its conceptual nodes into a multidimensional space of layer attributes and differentiating between an intensional and a preextensional level within the knowledge representation itself. The paper gives an overview of the expressional means of MultiNet and their use for representing linguistic knowledge and world knowledge.